YOLOv3 Object Detection

05/31/2019

YOLOv3 Object Detector in Keras with TensorFlow backend. This notebook uses a YOLOv3 network, pre-trained on the ImageNet and Microsoft COCO datasets. OpenCV (CV2) used for reading and saving images.

Using libraries YOLOv3 from

YOLOv3 Concepts

Image Grid

Anchor boxes

Bounding boxes

Training data

In [1]:
# Using TensorFlow 1.13.1
# YOLOv3 Keras code for TF 1.x; not compatible with TF 2.0 alpha

import numpy as np
import os
import matplotlib.pyplot as plt
from datetime import datetime

# Install CV2 with `pip install opencv-python`, rather than Anaconda
import cv2

import tensorflow as tf
import tensorflow.keras as keras

print("TensorFlow:", tf.__version__, " Keras:", keras.__version__, " CV2:", cv2.__version__)

TensorFlow: 1.13.1  Keras: 2.2.4-tf  CV2: 4.1.0

Define and load the Keras YOLOv3 model (106 layers)

Import Keras model experiencor/keras-yolo3

Main functions used:

yolov3 = make_yolov3_model()
weight_reader = WeightReader(weights_path)
weight_reader.load_weights(yolov3)

It takes about 1 minute to load YOLO model (Intel i7 quad-core 2.8 GHz)

CPU times: user 54.4 s, sys: 1.97 s, total: 56.3 s
Wall time: 55.6 s
In [2]:
# Import experiencor/keras-yolo3 library (some code changes for error debugging)
from keras_yolo3 import yolo3_one_file_to_detect_them_all as xp_yolov3

# YOLOv3 main() #1 
# args.weights
weights_path = '/Users/nelson/dev/cv/yolov3-keras/models/yolov3.weights'
# args.image
image_path = './data/images/african_elephant_people.jpg'  # sample image

# make the yolov3 model to predict 80 classes on COCO
print("make_yolov3_model:")
yolov3 = xp_yolov3.make_yolov3_model()
print()

# define weights reader
weight_reader = xp_yolov3.WeightReader(weights_path)

make_yolov3_model:
WARNING:tensorflow:From /Users/nelson/dev/anaconda3/lib/python3.7/site-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.
Instructions for updating:
Colocations handled automatically by placer.

Using TensorFlow backend.





















YOLOv3 - load weights












In [3]:



    


# YOLOv3 - load the weights trained on MS-COCO into the model
print("load_weights(yolov3):")
%time weight_reader.load_weights(yolov3)



    















    






load_weights(yolov3):
loading weights of convolution #0
loading weights of convolution #1
loading weights of convolution #2
loading weights of convolution #3
no convolution #4
loading weights of convolution #5
loading weights of convolution #6
loading weights of convolution #7
no convolution #8
loading weights of convolution #9
loading weights of convolution #10
no convolution #11
loading weights of convolution #12
loading weights of convolution #13
loading weights of convolution #14
no convolution #15
loading weights of convolution #16
loading weights of convolution #17
no convolution #18
loading weights of convolution #19
loading weights of convolution #20
no convolution #21
loading weights of convolution #22
loading weights of convolution #23
no convolution #24
loading weights of convolution #25
loading weights of convolution #26
no convolution #27
loading weights of convolution #28
loading weights of convolution #29
no convolution #30
loading weights of convolution #31
loading weights of convolution #32
no convolution #33
loading weights of convolution #34
loading weights of convolution #35
no convolution #36
loading weights of convolution #37
loading weights of convolution #38
loading weights of convolution #39
no convolution #40
loading weights of convolution #41
loading weights of convolution #42
no convolution #43
loading weights of convolution #44
loading weights of convolution #45
no convolution #46
loading weights of convolution #47
loading weights of convolution #48
no convolution #49
loading weights of convolution #50
loading weights of convolution #51
no convolution #52
loading weights of convolution #53
loading weights of convolution #54
no convolution #55
loading weights of convolution #56
loading weights of convolution #57
no convolution #58
loading weights of convolution #59
loading weights of convolution #60
no convolution #61
loading weights of convolution #62
loading weights of convolution #63
loading weights of convolution #64
no convolution #65
loading weights of convolution #66
loading weights of convolution #67
no convolution #68
loading weights of convolution #69
loading weights of convolution #70
no convolution #71
loading weights of convolution #72
loading weights of convolution #73
no convolution #74
loading weights of convolution #75
loading weights of convolution #76
loading weights of convolution #77
loading weights of convolution #78
loading weights of convolution #79
loading weights of convolution #80
loading weights of convolution #81
no convolution #82
no convolution #83
loading weights of convolution #84
no convolution #85
no convolution #86
loading weights of convolution #87
loading weights of convolution #88
loading weights of convolution #89
loading weights of convolution #90
loading weights of convolution #91
loading weights of convolution #92
loading weights of convolution #93
no convolution #94
no convolution #95
loading weights of convolution #96
no convolution #97
no convolution #98
loading weights of convolution #99
loading weights of convolution #100
loading weights of convolution #101
loading weights of convolution #102
loading weights of convolution #103
loading weights of convolution #104
loading weights of convolution #105
CPU times: user 54.4 s, sys: 2.04 s, total: 56.4 s
Wall time: 55.7 s



















In [4]:



    


# <tensorflow.python.keras.engine.training.Model at 0x14caaf898>
print ("Number of yolov3.layers:", len(yolov3.layers))



    















    






Number of yolov3.layers: 252



















In [5]:



    


# YOLOv3 Keras model summary
yolov3.summary()



    















    






__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_1 (InputLayer)            (None, None, None, 3 0                                            
__________________________________________________________________________________________________
conv_0 (Conv2D)                 (None, None, None, 3 864         input_1[0][0]                    
__________________________________________________________________________________________________
bnorm_0 (BatchNormalization)    (None, None, None, 3 128         conv_0[0][0]                     
__________________________________________________________________________________________________
leaky_0 (LeakyReLU)             (None, None, None, 3 0           bnorm_0[0][0]                    
__________________________________________________________________________________________________
zero_padding2d_1 (ZeroPadding2D (None, None, None, 3 0           leaky_0[0][0]                    
__________________________________________________________________________________________________
conv_1 (Conv2D)                 (None, None, None, 6 18432       zero_padding2d_1[0][0]           
__________________________________________________________________________________________________
bnorm_1 (BatchNormalization)    (None, None, None, 6 256         conv_1[0][0]                     
__________________________________________________________________________________________________
leaky_1 (LeakyReLU)             (None, None, None, 6 0           bnorm_1[0][0]                    
__________________________________________________________________________________________________
conv_2 (Conv2D)                 (None, None, None, 3 2048        leaky_1[0][0]                    
__________________________________________________________________________________________________
bnorm_2 (BatchNormalization)    (None, None, None, 3 128         conv_2[0][0]                     
__________________________________________________________________________________________________
leaky_2 (LeakyReLU)             (None, None, None, 3 0           bnorm_2[0][0]                    
__________________________________________________________________________________________________
conv_3 (Conv2D)                 (None, None, None, 6 18432       leaky_2[0][0]                    
__________________________________________________________________________________________________
bnorm_3 (BatchNormalization)    (None, None, None, 6 256         conv_3[0][0]                     
__________________________________________________________________________________________________
leaky_3 (LeakyReLU)             (None, None, None, 6 0           bnorm_3[0][0]                    
__________________________________________________________________________________________________
add_1 (Add)                     (None, None, None, 6 0           leaky_1[0][0]                    
                                                                 leaky_3[0][0]                    
__________________________________________________________________________________________________
zero_padding2d_2 (ZeroPadding2D (None, None, None, 6 0           add_1[0][0]                      
__________________________________________________________________________________________________
conv_5 (Conv2D)                 (None, None, None, 1 73728       zero_padding2d_2[0][0]           
__________________________________________________________________________________________________
bnorm_5 (BatchNormalization)    (None, None, None, 1 512         conv_5[0][0]                     
__________________________________________________________________________________________________
leaky_5 (LeakyReLU)             (None, None, None, 1 0           bnorm_5[0][0]                    
__________________________________________________________________________________________________
conv_6 (Conv2D)                 (None, None, None, 6 8192        leaky_5[0][0]                    
__________________________________________________________________________________________________
bnorm_6 (BatchNormalization)    (None, None, None, 6 256         conv_6[0][0]                     
__________________________________________________________________________________________________
leaky_6 (LeakyReLU)             (None, None, None, 6 0           bnorm_6[0][0]                    
__________________________________________________________________________________________________
conv_7 (Conv2D)                 (None, None, None, 1 73728       leaky_6[0][0]                    
__________________________________________________________________________________________________
bnorm_7 (BatchNormalization)    (None, None, None, 1 512         conv_7[0][0]                     
__________________________________________________________________________________________________
leaky_7 (LeakyReLU)             (None, None, None, 1 0           bnorm_7[0][0]                    
__________________________________________________________________________________________________
add_2 (Add)                     (None, None, None, 1 0           leaky_5[0][0]                    
                                                                 leaky_7[0][0]                    
__________________________________________________________________________________________________
conv_9 (Conv2D)                 (None, None, None, 6 8192        add_2[0][0]                      
__________________________________________________________________________________________________
bnorm_9 (BatchNormalization)    (None, None, None, 6 256         conv_9[0][0]                     
__________________________________________________________________________________________________
leaky_9 (LeakyReLU)             (None, None, None, 6 0           bnorm_9[0][0]                    
__________________________________________________________________________________________________
conv_10 (Conv2D)                (None, None, None, 1 73728       leaky_9[0][0]                    
__________________________________________________________________________________________________
bnorm_10 (BatchNormalization)   (None, None, None, 1 512         conv_10[0][0]                    
__________________________________________________________________________________________________
leaky_10 (LeakyReLU)            (None, None, None, 1 0           bnorm_10[0][0]                   
__________________________________________________________________________________________________
add_3 (Add)                     (None, None, None, 1 0           add_2[0][0]                      
                                                                 leaky_10[0][0]                   
__________________________________________________________________________________________________
zero_padding2d_3 (ZeroPadding2D (None, None, None, 1 0           add_3[0][0]                      
__________________________________________________________________________________________________
conv_12 (Conv2D)                (None, None, None, 2 294912      zero_padding2d_3[0][0]           
__________________________________________________________________________________________________
bnorm_12 (BatchNormalization)   (None, None, None, 2 1024        conv_12[0][0]                    
__________________________________________________________________________________________________
leaky_12 (LeakyReLU)            (None, None, None, 2 0           bnorm_12[0][0]                   
__________________________________________________________________________________________________
conv_13 (Conv2D)                (None, None, None, 1 32768       leaky_12[0][0]                   
__________________________________________________________________________________________________
bnorm_13 (BatchNormalization)   (None, None, None, 1 512         conv_13[0][0]                    
__________________________________________________________________________________________________
leaky_13 (LeakyReLU)            (None, None, None, 1 0           bnorm_13[0][0]                   
__________________________________________________________________________________________________
conv_14 (Conv2D)                (None, None, None, 2 294912      leaky_13[0][0]                   
__________________________________________________________________________________________________
bnorm_14 (BatchNormalization)   (None, None, None, 2 1024        conv_14[0][0]                    
__________________________________________________________________________________________________
leaky_14 (LeakyReLU)            (None, None, None, 2 0           bnorm_14[0][0]                   
__________________________________________________________________________________________________
add_4 (Add)                     (None, None, None, 2 0           leaky_12[0][0]                   
                                                                 leaky_14[0][0]                   
__________________________________________________________________________________________________
conv_16 (Conv2D)                (None, None, None, 1 32768       add_4[0][0]                      
__________________________________________________________________________________________________
bnorm_16 (BatchNormalization)   (None, None, None, 1 512         conv_16[0][0]                    
__________________________________________________________________________________________________
leaky_16 (LeakyReLU)            (None, None, None, 1 0           bnorm_16[0][0]                   
__________________________________________________________________________________________________
conv_17 (Conv2D)                (None, None, None, 2 294912      leaky_16[0][0]                   
__________________________________________________________________________________________________
bnorm_17 (BatchNormalization)   (None, None, None, 2 1024        conv_17[0][0]                    
__________________________________________________________________________________________________
leaky_17 (LeakyReLU)            (None, None, None, 2 0           bnorm_17[0][0]                   
__________________________________________________________________________________________________
add_5 (Add)                     (None, None, None, 2 0           add_4[0][0]                      
                                                                 leaky_17[0][0]                   
__________________________________________________________________________________________________
conv_19 (Conv2D)                (None, None, None, 1 32768       add_5[0][0]                      
__________________________________________________________________________________________________
bnorm_19 (BatchNormalization)   (None, None, None, 1 512         conv_19[0][0]                    
__________________________________________________________________________________________________
leaky_19 (LeakyReLU)            (None, None, None, 1 0           bnorm_19[0][0]                   
__________________________________________________________________________________________________
conv_20 (Conv2D)                (None, None, None, 2 294912      leaky_19[0][0]                   
__________________________________________________________________________________________________
bnorm_20 (BatchNormalization)   (None, None, None, 2 1024        conv_20[0][0]                    
__________________________________________________________________________________________________
leaky_20 (LeakyReLU)            (None, None, None, 2 0           bnorm_20[0][0]                   
__________________________________________________________________________________________________
add_6 (Add)                     (None, None, None, 2 0           add_5[0][0]                      
                                                                 leaky_20[0][0]                   
__________________________________________________________________________________________________
conv_22 (Conv2D)                (None, None, None, 1 32768       add_6[0][0]                      
__________________________________________________________________________________________________
bnorm_22 (BatchNormalization)   (None, None, None, 1 512         conv_22[0][0]                    
__________________________________________________________________________________________________
leaky_22 (LeakyReLU)            (None, None, None, 1 0           bnorm_22[0][0]                   
__________________________________________________________________________________________________
conv_23 (Conv2D)                (None, None, None, 2 294912      leaky_22[0][0]                   
__________________________________________________________________________________________________
bnorm_23 (BatchNormalization)   (None, None, None, 2 1024        conv_23[0][0]                    
__________________________________________________________________________________________________
leaky_23 (LeakyReLU)            (None, None, None, 2 0           bnorm_23[0][0]                   
__________________________________________________________________________________________________
add_7 (Add)                     (None, None, None, 2 0           add_6[0][0]                      
                                                                 leaky_23[0][0]                   
__________________________________________________________________________________________________
conv_25 (Conv2D)                (None, None, None, 1 32768       add_7[0][0]                      
__________________________________________________________________________________________________
bnorm_25 (BatchNormalization)   (None, None, None, 1 512         conv_25[0][0]                    
__________________________________________________________________________________________________
leaky_25 (LeakyReLU)            (None, None, None, 1 0           bnorm_25[0][0]                   
__________________________________________________________________________________________________
conv_26 (Conv2D)                (None, None, None, 2 294912      leaky_25[0][0]                   
__________________________________________________________________________________________________
bnorm_26 (BatchNormalization)   (None, None, None, 2 1024        conv_26[0][0]                    
__________________________________________________________________________________________________
leaky_26 (LeakyReLU)            (None, None, None, 2 0           bnorm_26[0][0]                   
__________________________________________________________________________________________________
add_8 (Add)                     (None, None, None, 2 0           add_7[0][0]                      
                                                                 leaky_26[0][0]                   
__________________________________________________________________________________________________
conv_28 (Conv2D)                (None, None, None, 1 32768       add_8[0][0]                      
__________________________________________________________________________________________________
bnorm_28 (BatchNormalization)   (None, None, None, 1 512         conv_28[0][0]                    
__________________________________________________________________________________________________
leaky_28 (LeakyReLU)            (None, None, None, 1 0           bnorm_28[0][0]                   
__________________________________________________________________________________________________
conv_29 (Conv2D)                (None, None, None, 2 294912      leaky_28[0][0]                   
__________________________________________________________________________________________________
bnorm_29 (BatchNormalization)   (None, None, None, 2 1024        conv_29[0][0]                    
__________________________________________________________________________________________________
leaky_29 (LeakyReLU)            (None, None, None, 2 0           bnorm_29[0][0]                   
__________________________________________________________________________________________________
add_9 (Add)                     (None, None, None, 2 0           add_8[0][0]                      
                                                                 leaky_29[0][0]                   
__________________________________________________________________________________________________
conv_31 (Conv2D)                (None, None, None, 1 32768       add_9[0][0]                      
__________________________________________________________________________________________________
bnorm_31 (BatchNormalization)   (None, None, None, 1 512         conv_31[0][0]                    
__________________________________________________________________________________________________
leaky_31 (LeakyReLU)            (None, None, None, 1 0           bnorm_31[0][0]                   
__________________________________________________________________________________________________
conv_32 (Conv2D)                (None, None, None, 2 294912      leaky_31[0][0]                   
__________________________________________________________________________________________________
bnorm_32 (BatchNormalization)   (None, None, None, 2 1024        conv_32[0][0]                    
__________________________________________________________________________________________________
leaky_32 (LeakyReLU)            (None, None, None, 2 0           bnorm_32[0][0]                   
__________________________________________________________________________________________________
add_10 (Add)                    (None, None, None, 2 0           add_9[0][0]                      
                                                                 leaky_32[0][0]                   
__________________________________________________________________________________________________
conv_34 (Conv2D)                (None, None, None, 1 32768       add_10[0][0]                     
__________________________________________________________________________________________________
bnorm_34 (BatchNormalization)   (None, None, None, 1 512         conv_34[0][0]                    
__________________________________________________________________________________________________
leaky_34 (LeakyReLU)            (None, None, None, 1 0           bnorm_34[0][0]                   
__________________________________________________________________________________________________
conv_35 (Conv2D)                (None, None, None, 2 294912      leaky_34[0][0]                   
__________________________________________________________________________________________________
bnorm_35 (BatchNormalization)   (None, None, None, 2 1024        conv_35[0][0]                    
__________________________________________________________________________________________________
leaky_35 (LeakyReLU)            (None, None, None, 2 0           bnorm_35[0][0]                   
__________________________________________________________________________________________________
add_11 (Add)                    (None, None, None, 2 0           add_10[0][0]                     
                                                                 leaky_35[0][0]                   
__________________________________________________________________________________________________
zero_padding2d_4 (ZeroPadding2D (None, None, None, 2 0           add_11[0][0]                     
__________________________________________________________________________________________________
conv_37 (Conv2D)                (None, None, None, 5 1179648     zero_padding2d_4[0][0]           
__________________________________________________________________________________________________
bnorm_37 (BatchNormalization)   (None, None, None, 5 2048        conv_37[0][0]                    
__________________________________________________________________________________________________
leaky_37 (LeakyReLU)            (None, None, None, 5 0           bnorm_37[0][0]                   
__________________________________________________________________________________________________
conv_38 (Conv2D)                (None, None, None, 2 131072      leaky_37[0][0]                   
__________________________________________________________________________________________________
bnorm_38 (BatchNormalization)   (None, None, None, 2 1024        conv_38[0][0]                    
__________________________________________________________________________________________________
leaky_38 (LeakyReLU)            (None, None, None, 2 0           bnorm_38[0][0]                   
__________________________________________________________________________________________________
conv_39 (Conv2D)                (None, None, None, 5 1179648     leaky_38[0][0]                   
__________________________________________________________________________________________________
bnorm_39 (BatchNormalization)   (None, None, None, 5 2048        conv_39[0][0]                    
__________________________________________________________________________________________________
leaky_39 (LeakyReLU)            (None, None, None, 5 0           bnorm_39[0][0]                   
__________________________________________________________________________________________________
add_12 (Add)                    (None, None, None, 5 0           leaky_37[0][0]                   
                                                                 leaky_39[0][0]                   
__________________________________________________________________________________________________
conv_41 (Conv2D)                (None, None, None, 2 131072      add_12[0][0]                     
__________________________________________________________________________________________________
bnorm_41 (BatchNormalization)   (None, None, None, 2 1024        conv_41[0][0]                    
__________________________________________________________________________________________________
leaky_41 (LeakyReLU)            (None, None, None, 2 0           bnorm_41[0][0]                   
__________________________________________________________________________________________________
conv_42 (Conv2D)                (None, None, None, 5 1179648     leaky_41[0][0]                   
__________________________________________________________________________________________________
bnorm_42 (BatchNormalization)   (None, None, None, 5 2048        conv_42[0][0]                    
__________________________________________________________________________________________________
leaky_42 (LeakyReLU)            (None, None, None, 5 0           bnorm_42[0][0]                   
__________________________________________________________________________________________________
add_13 (Add)                    (None, None, None, 5 0           add_12[0][0]                     
                                                                 leaky_42[0][0]                   
__________________________________________________________________________________________________
conv_44 (Conv2D)                (None, None, None, 2 131072      add_13[0][0]                     
__________________________________________________________________________________________________
bnorm_44 (BatchNormalization)   (None, None, None, 2 1024        conv_44[0][0]                    
__________________________________________________________________________________________________
leaky_44 (LeakyReLU)            (None, None, None, 2 0           bnorm_44[0][0]                   
__________________________________________________________________________________________________
conv_45 (Conv2D)                (None, None, None, 5 1179648     leaky_44[0][0]                   
__________________________________________________________________________________________________
bnorm_45 (BatchNormalization)   (None, None, None, 5 2048        conv_45[0][0]                    
__________________________________________________________________________________________________
leaky_45 (LeakyReLU)            (None, None, None, 5 0           bnorm_45[0][0]                   
__________________________________________________________________________________________________
add_14 (Add)                    (None, None, None, 5 0           add_13[0][0]                     
                                                                 leaky_45[0][0]                   
__________________________________________________________________________________________________
conv_47 (Conv2D)                (None, None, None, 2 131072      add_14[0][0]                     
__________________________________________________________________________________________________
bnorm_47 (BatchNormalization)   (None, None, None, 2 1024        conv_47[0][0]                    
__________________________________________________________________________________________________
leaky_47 (LeakyReLU)            (None, None, None, 2 0           bnorm_47[0][0]                   
__________________________________________________________________________________________________
conv_48 (Conv2D)                (None, None, None, 5 1179648     leaky_47[0][0]                   
__________________________________________________________________________________________________
bnorm_48 (BatchNormalization)   (None, None, None, 5 2048        conv_48[0][0]                    
__________________________________________________________________________________________________
leaky_48 (LeakyReLU)            (None, None, None, 5 0           bnorm_48[0][0]                   
__________________________________________________________________________________________________
add_15 (Add)                    (None, None, None, 5 0           add_14[0][0]                     
                                                                 leaky_48[0][0]                   
__________________________________________________________________________________________________
conv_50 (Conv2D)                (None, None, None, 2 131072      add_15[0][0]                     
__________________________________________________________________________________________________
bnorm_50 (BatchNormalization)   (None, None, None, 2 1024        conv_50[0][0]                    
__________________________________________________________________________________________________
leaky_50 (LeakyReLU)            (None, None, None, 2 0           bnorm_50[0][0]                   
__________________________________________________________________________________________________
conv_51 (Conv2D)                (None, None, None, 5 1179648     leaky_50[0][0]                   
__________________________________________________________________________________________________
bnorm_51 (BatchNormalization)   (None, None, None, 5 2048        conv_51[0][0]                    
__________________________________________________________________________________________________
leaky_51 (LeakyReLU)            (None, None, None, 5 0           bnorm_51[0][0]                   
__________________________________________________________________________________________________
add_16 (Add)                    (None, None, None, 5 0           add_15[0][0]                     
                                                                 leaky_51[0][0]                   
__________________________________________________________________________________________________
conv_53 (Conv2D)                (None, None, None, 2 131072      add_16[0][0]                     
__________________________________________________________________________________________________
bnorm_53 (BatchNormalization)   (None, None, None, 2 1024        conv_53[0][0]                    
__________________________________________________________________________________________________
leaky_53 (LeakyReLU)            (None, None, None, 2 0           bnorm_53[0][0]                   
__________________________________________________________________________________________________
conv_54 (Conv2D)                (None, None, None, 5 1179648     leaky_53[0][0]                   
__________________________________________________________________________________________________
bnorm_54 (BatchNormalization)   (None, None, None, 5 2048        conv_54[0][0]                    
__________________________________________________________________________________________________
leaky_54 (LeakyReLU)            (None, None, None, 5 0           bnorm_54[0][0]                   
__________________________________________________________________________________________________
add_17 (Add)                    (None, None, None, 5 0           add_16[0][0]                     
                                                                 leaky_54[0][0]                   
__________________________________________________________________________________________________
conv_56 (Conv2D)                (None, None, None, 2 131072      add_17[0][0]                     
__________________________________________________________________________________________________
bnorm_56 (BatchNormalization)   (None, None, None, 2 1024        conv_56[0][0]                    
__________________________________________________________________________________________________
leaky_56 (LeakyReLU)            (None, None, None, 2 0           bnorm_56[0][0]                   
__________________________________________________________________________________________________
conv_57 (Conv2D)                (None, None, None, 5 1179648     leaky_56[0][0]                   
__________________________________________________________________________________________________
bnorm_57 (BatchNormalization)   (None, None, None, 5 2048        conv_57[0][0]                    
__________________________________________________________________________________________________
leaky_57 (LeakyReLU)            (None, None, None, 5 0           bnorm_57[0][0]                   
__________________________________________________________________________________________________
add_18 (Add)                    (None, None, None, 5 0           add_17[0][0]                     
                                                                 leaky_57[0][0]                   
__________________________________________________________________________________________________
conv_59 (Conv2D)                (None, None, None, 2 131072      add_18[0][0]                     
__________________________________________________________________________________________________
bnorm_59 (BatchNormalization)   (None, None, None, 2 1024        conv_59[0][0]                    
__________________________________________________________________________________________________
leaky_59 (LeakyReLU)            (None, None, None, 2 0           bnorm_59[0][0]                   
__________________________________________________________________________________________________
conv_60 (Conv2D)                (None, None, None, 5 1179648     leaky_59[0][0]                   
__________________________________________________________________________________________________
bnorm_60 (BatchNormalization)   (None, None, None, 5 2048        conv_60[0][0]                    
__________________________________________________________________________________________________
leaky_60 (LeakyReLU)            (None, None, None, 5 0           bnorm_60[0][0]                   
__________________________________________________________________________________________________
add_19 (Add)                    (None, None, None, 5 0           add_18[0][0]                     
                                                                 leaky_60[0][0]                   
__________________________________________________________________________________________________
zero_padding2d_5 (ZeroPadding2D (None, None, None, 5 0           add_19[0][0]                     
__________________________________________________________________________________________________
conv_62 (Conv2D)                (None, None, None, 1 4718592     zero_padding2d_5[0][0]           
__________________________________________________________________________________________________
bnorm_62 (BatchNormalization)   (None, None, None, 1 4096        conv_62[0][0]                    
__________________________________________________________________________________________________
leaky_62 (LeakyReLU)            (None, None, None, 1 0           bnorm_62[0][0]                   
__________________________________________________________________________________________________
conv_63 (Conv2D)                (None, None, None, 5 524288      leaky_62[0][0]                   
__________________________________________________________________________________________________
bnorm_63 (BatchNormalization)   (None, None, None, 5 2048        conv_63[0][0]                    
__________________________________________________________________________________________________
leaky_63 (LeakyReLU)            (None, None, None, 5 0           bnorm_63[0][0]                   
__________________________________________________________________________________________________
conv_64 (Conv2D)                (None, None, None, 1 4718592     leaky_63[0][0]                   
__________________________________________________________________________________________________
bnorm_64 (BatchNormalization)   (None, None, None, 1 4096        conv_64[0][0]                    
__________________________________________________________________________________________________
leaky_64 (LeakyReLU)            (None, None, None, 1 0           bnorm_64[0][0]                   
__________________________________________________________________________________________________
add_20 (Add)                    (None, None, None, 1 0           leaky_62[0][0]                   
                                                                 leaky_64[0][0]                   
__________________________________________________________________________________________________
conv_66 (Conv2D)                (None, None, None, 5 524288      add_20[0][0]                     
__________________________________________________________________________________________________
bnorm_66 (BatchNormalization)   (None, None, None, 5 2048        conv_66[0][0]                    
__________________________________________________________________________________________________
leaky_66 (LeakyReLU)            (None, None, None, 5 0           bnorm_66[0][0]                   
__________________________________________________________________________________________________
conv_67 (Conv2D)                (None, None, None, 1 4718592     leaky_66[0][0]                   
__________________________________________________________________________________________________
bnorm_67 (BatchNormalization)   (None, None, None, 1 4096        conv_67[0][0]                    
__________________________________________________________________________________________________
leaky_67 (LeakyReLU)            (None, None, None, 1 0           bnorm_67[0][0]                   
__________________________________________________________________________________________________
add_21 (Add)                    (None, None, None, 1 0           add_20[0][0]                     
                                                                 leaky_67[0][0]                   
__________________________________________________________________________________________________
conv_69 (Conv2D)                (None, None, None, 5 524288      add_21[0][0]                     
__________________________________________________________________________________________________
bnorm_69 (BatchNormalization)   (None, None, None, 5 2048        conv_69[0][0]                    
__________________________________________________________________________________________________
leaky_69 (LeakyReLU)            (None, None, None, 5 0           bnorm_69[0][0]                   
__________________________________________________________________________________________________
conv_70 (Conv2D)                (None, None, None, 1 4718592     leaky_69[0][0]                   
__________________________________________________________________________________________________
bnorm_70 (BatchNormalization)   (None, None, None, 1 4096        conv_70[0][0]                    
__________________________________________________________________________________________________
leaky_70 (LeakyReLU)            (None, None, None, 1 0           bnorm_70[0][0]                   
__________________________________________________________________________________________________
add_22 (Add)                    (None, None, None, 1 0           add_21[0][0]                     
                                                                 leaky_70[0][0]                   
__________________________________________________________________________________________________
conv_72 (Conv2D)                (None, None, None, 5 524288      add_22[0][0]                     
__________________________________________________________________________________________________
bnorm_72 (BatchNormalization)   (None, None, None, 5 2048        conv_72[0][0]                    
__________________________________________________________________________________________________
leaky_72 (LeakyReLU)            (None, None, None, 5 0           bnorm_72[0][0]                   
__________________________________________________________________________________________________
conv_73 (Conv2D)                (None, None, None, 1 4718592     leaky_72[0][0]                   
__________________________________________________________________________________________________
bnorm_73 (BatchNormalization)   (None, None, None, 1 4096        conv_73[0][0]                    
__________________________________________________________________________________________________
leaky_73 (LeakyReLU)            (None, None, None, 1 0           bnorm_73[0][0]                   
__________________________________________________________________________________________________
add_23 (Add)                    (None, None, None, 1 0           add_22[0][0]                     
                                                                 leaky_73[0][0]                   
__________________________________________________________________________________________________
conv_75 (Conv2D)                (None, None, None, 5 524288      add_23[0][0]                     
__________________________________________________________________________________________________
bnorm_75 (BatchNormalization)   (None, None, None, 5 2048        conv_75[0][0]                    
__________________________________________________________________________________________________
leaky_75 (LeakyReLU)            (None, None, None, 5 0           bnorm_75[0][0]                   
__________________________________________________________________________________________________
conv_76 (Conv2D)                (None, None, None, 1 4718592     leaky_75[0][0]                   
__________________________________________________________________________________________________
bnorm_76 (BatchNormalization)   (None, None, None, 1 4096        conv_76[0][0]                    
__________________________________________________________________________________________________
leaky_76 (LeakyReLU)            (None, None, None, 1 0           bnorm_76[0][0]                   
__________________________________________________________________________________________________
conv_77 (Conv2D)                (None, None, None, 5 524288      leaky_76[0][0]                   
__________________________________________________________________________________________________
bnorm_77 (BatchNormalization)   (None, None, None, 5 2048        conv_77[0][0]                    
__________________________________________________________________________________________________
leaky_77 (LeakyReLU)            (None, None, None, 5 0           bnorm_77[0][0]                   
__________________________________________________________________________________________________
conv_78 (Conv2D)                (None, None, None, 1 4718592     leaky_77[0][0]                   
__________________________________________________________________________________________________
bnorm_78 (BatchNormalization)   (None, None, None, 1 4096        conv_78[0][0]                    
__________________________________________________________________________________________________
leaky_78 (LeakyReLU)            (None, None, None, 1 0           bnorm_78[0][0]                   
__________________________________________________________________________________________________
conv_79 (Conv2D)                (None, None, None, 5 524288      leaky_78[0][0]                   
__________________________________________________________________________________________________
bnorm_79 (BatchNormalization)   (None, None, None, 5 2048        conv_79[0][0]                    
__________________________________________________________________________________________________
leaky_79 (LeakyReLU)            (None, None, None, 5 0           bnorm_79[0][0]                   
__________________________________________________________________________________________________
conv_84 (Conv2D)                (None, None, None, 2 131072      leaky_79[0][0]                   
__________________________________________________________________________________________________
bnorm_84 (BatchNormalization)   (None, None, None, 2 1024        conv_84[0][0]                    
__________________________________________________________________________________________________
leaky_84 (LeakyReLU)            (None, None, None, 2 0           bnorm_84[0][0]                   
__________________________________________________________________________________________________
up_sampling2d_1 (UpSampling2D)  (None, None, None, 2 0           leaky_84[0][0]                   
__________________________________________________________________________________________________
concatenate_1 (Concatenate)     (None, None, None, 7 0           up_sampling2d_1[0][0]            
                                                                 add_19[0][0]                     
__________________________________________________________________________________________________
conv_87 (Conv2D)                (None, None, None, 2 196608      concatenate_1[0][0]              
__________________________________________________________________________________________________
bnorm_87 (BatchNormalization)   (None, None, None, 2 1024        conv_87[0][0]                    
__________________________________________________________________________________________________
leaky_87 (LeakyReLU)            (None, None, None, 2 0           bnorm_87[0][0]                   
__________________________________________________________________________________________________
conv_88 (Conv2D)                (None, None, None, 5 1179648     leaky_87[0][0]                   
__________________________________________________________________________________________________
bnorm_88 (BatchNormalization)   (None, None, None, 5 2048        conv_88[0][0]                    
__________________________________________________________________________________________________
leaky_88 (LeakyReLU)            (None, None, None, 5 0           bnorm_88[0][0]                   
__________________________________________________________________________________________________
conv_89 (Conv2D)                (None, None, None, 2 131072      leaky_88[0][0]                   
__________________________________________________________________________________________________
bnorm_89 (BatchNormalization)   (None, None, None, 2 1024        conv_89[0][0]                    
__________________________________________________________________________________________________
leaky_89 (LeakyReLU)            (None, None, None, 2 0           bnorm_89[0][0]                   
__________________________________________________________________________________________________
conv_90 (Conv2D)                (None, None, None, 5 1179648     leaky_89[0][0]                   
__________________________________________________________________________________________________
bnorm_90 (BatchNormalization)   (None, None, None, 5 2048        conv_90[0][0]                    
__________________________________________________________________________________________________
leaky_90 (LeakyReLU)            (None, None, None, 5 0           bnorm_90[0][0]                   
__________________________________________________________________________________________________
conv_91 (Conv2D)                (None, None, None, 2 131072      leaky_90[0][0]                   
__________________________________________________________________________________________________
bnorm_91 (BatchNormalization)   (None, None, None, 2 1024        conv_91[0][0]                    
__________________________________________________________________________________________________
leaky_91 (LeakyReLU)            (None, None, None, 2 0           bnorm_91[0][0]                   
__________________________________________________________________________________________________
conv_96 (Conv2D)                (None, None, None, 1 32768       leaky_91[0][0]                   
__________________________________________________________________________________________________
bnorm_96 (BatchNormalization)   (None, None, None, 1 512         conv_96[0][0]                    
__________________________________________________________________________________________________
leaky_96 (LeakyReLU)            (None, None, None, 1 0           bnorm_96[0][0]                   
__________________________________________________________________________________________________
up_sampling2d_2 (UpSampling2D)  (None, None, None, 1 0           leaky_96[0][0]                   
__________________________________________________________________________________________________
concatenate_2 (Concatenate)     (None, None, None, 3 0           up_sampling2d_2[0][0]            
                                                                 add_11[0][0]                     
__________________________________________________________________________________________________
conv_99 (Conv2D)                (None, None, None, 1 49152       concatenate_2[0][0]              
__________________________________________________________________________________________________
bnorm_99 (BatchNormalization)   (None, None, None, 1 512         conv_99[0][0]                    
__________________________________________________________________________________________________
leaky_99 (LeakyReLU)            (None, None, None, 1 0           bnorm_99[0][0]                   
__________________________________________________________________________________________________
conv_100 (Conv2D)               (None, None, None, 2 294912      leaky_99[0][0]                   
__________________________________________________________________________________________________
bnorm_100 (BatchNormalization)  (None, None, None, 2 1024        conv_100[0][0]                   
__________________________________________________________________________________________________
leaky_100 (LeakyReLU)           (None, None, None, 2 0           bnorm_100[0][0]                  
__________________________________________________________________________________________________
conv_101 (Conv2D)               (None, None, None, 1 32768       leaky_100[0][0]                  
__________________________________________________________________________________________________
bnorm_101 (BatchNormalization)  (None, None, None, 1 512         conv_101[0][0]                   
__________________________________________________________________________________________________
leaky_101 (LeakyReLU)           (None, None, None, 1 0           bnorm_101[0][0]                  
__________________________________________________________________________________________________
conv_102 (Conv2D)               (None, None, None, 2 294912      leaky_101[0][0]                  
__________________________________________________________________________________________________
bnorm_102 (BatchNormalization)  (None, None, None, 2 1024        conv_102[0][0]                   
__________________________________________________________________________________________________
leaky_102 (LeakyReLU)           (None, None, None, 2 0           bnorm_102[0][0]                  
__________________________________________________________________________________________________
conv_103 (Conv2D)               (None, None, None, 1 32768       leaky_102[0][0]                  
__________________________________________________________________________________________________
bnorm_103 (BatchNormalization)  (None, None, None, 1 512         conv_103[0][0]                   
__________________________________________________________________________________________________
leaky_103 (LeakyReLU)           (None, None, None, 1 0           bnorm_103[0][0]                  
__________________________________________________________________________________________________
conv_80 (Conv2D)                (None, None, None, 1 4718592     leaky_79[0][0]                   
__________________________________________________________________________________________________
conv_92 (Conv2D)                (None, None, None, 5 1179648     leaky_91[0][0]                   
__________________________________________________________________________________________________
conv_104 (Conv2D)               (None, None, None, 2 294912      leaky_103[0][0]                  
__________________________________________________________________________________________________
bnorm_80 (BatchNormalization)   (None, None, None, 1 4096        conv_80[0][0]                    
__________________________________________________________________________________________________
bnorm_92 (BatchNormalization)   (None, None, None, 5 2048        conv_92[0][0]                    
__________________________________________________________________________________________________
bnorm_104 (BatchNormalization)  (None, None, None, 2 1024        conv_104[0][0]                   
__________________________________________________________________________________________________
leaky_80 (LeakyReLU)            (None, None, None, 1 0           bnorm_80[0][0]                   
__________________________________________________________________________________________________
leaky_92 (LeakyReLU)            (None, None, None, 5 0           bnorm_92[0][0]                   
__________________________________________________________________________________________________
leaky_104 (LeakyReLU)           (None, None, None, 2 0           bnorm_104[0][0]                  
__________________________________________________________________________________________________
conv_81 (Conv2D)                (None, None, None, 2 261375      leaky_80[0][0]                   
__________________________________________________________________________________________________
conv_93 (Conv2D)                (None, None, None, 2 130815      leaky_92[0][0]                   
__________________________________________________________________________________________________
conv_105 (Conv2D)               (None, None, None, 2 65535       leaky_104[0][0]                  
==================================================================================================
Total params: 62,001,757
Trainable params: 61,949,149
Non-trainable params: 52,608
__________________________________________________________________________________________________





















YOLOv3 Notes


Total yolov3.layers: 252
Convolutional yolov3.layers: 106
Three image scales: layers 79-82 (stride 32), 91-94 (stride 16), 103-106 (stride 8)
Detection layer outputs: conv_81, conv_93, conv_105

Total params: 62,001,757
Trainable params: 61,949,149
Non-trainable params: 52,608





YOLOv3 Layers (0 - 105)
    

  • See YOLOv3 Keras model definition in ./keras_yolo3/yolo.py
    • 



Detection layer outputs


conv_81 (Conv2D)                (None, None, None, 2 261375      leaky_80[0][0]               conv_93 (Conv2D)                (None, None, None, 2 130815      leaky_92[0][0]               conv_105 (Conv2D)               (None, None, None, 2 65535       leaky_104[0][0]           















YOLOv3 Summary
Sample layers


<keras.engine.training.Model object at 0xb4a178a58>
Layer (type)                    Output Shape         Param #     Connected to
input_3 (InputLayer)            (None, None, None, 3 0 
conv_0 (Conv2D)                 (None, None, None, 3 864         input_3[0][0]   
bnorm_0 (BatchNormalization)    (None, None, None, 3 128         conv_0[0][0]
leaky_0 (LeakyReLU)             (None, None, None, 3 0           bnorm_0[0][0]
zero_padding2d_11 (ZeroPadding2 (None, None, None, 3 0           leaky_0[0][0]
...
add_47 (Add)                    (None, None, None, 6 0           leaky_1[0][0]                                                                                  leaky_3[0][0]        
zero_padding2d_12 (ZeroPadding2 (None, None, None, 6 0           add_47[0][0]           
conv_5 (Conv2D)                 (None, None, None, 1 73728       zero_padding2d_12[0][0]  
bnorm_5 (BatchNormalization)    (None, None, None, 1 512         conv_5[0][0]            
leaky_5 (LeakyReLU)             (None, None, None, 1 0           bnorm_5[0][0]           
...
conv_84 (Conv2D)                (None, None, None, 2 131072      leaky_79[0][0]               bnorm_84 (BatchNormalization)   (None, None, None, 2 1024        conv_84[0][0]                 leaky_84 (LeakyReLU)            (None, None, None, 2 0           bnorm_84[0][0]               up_sampling2d_5 (UpSampling2D)  (None, None, None, 2 0           leaky_84[0][0]           
concatenate_5 (Concatenate)     (None, None, None, 7 0           up_sampling2d_5[0][0]                                                                          add_65[0][0]                 ...
conv_96 (Conv2D)                (None, None, None, 1 32768       leaky_91[0][0]               bnorm_96 (BatchNormalization)   (None, None, None, 1 512         conv_96[0][0]                 leaky_96 (LeakyReLU)            (None, None, None, 1 0           bnorm_96[0][0]               up_sampling2d_6 (UpSampling2D)  (None, None, None, 1 0           leaky_96[0][0]               concatenate_6 (Concatenate)     (None, None, None, 3 0           up_sampling2d_6[0][0]                                                                          add_57[0][0]                 ...
conv_102 (Conv2D)               (None, None, None, 2 294912      leaky_101[0][0]               bnorm_102 (BatchNormalization)  (None, None, None, 2 1024        conv_102[0][0]               leaky_102 (LeakyReLU)           (None, None, None, 2 0           bnorm_102[0][0]               conv_103 (Conv2D)               (None, None, None, 1 32768       leaky_102[0][0]               bnorm_103 (BatchNormalization)  (None, None, None, 1 512         conv_103[0][0]               leaky_103 (LeakyReLU)           (None, None, None, 1 0           bnorm_103[0][0]               conv_80 (Conv2D)                (None, None, None, 1 4718592     leaky_79[0][0]               conv_92 (Conv2D)                (None, None, None, 5 1179648     leaky_91[0][0]               conv_104 (Conv2D)               (None, None, None, 2 294912      leaky_103[0][0]               bnorm_80 (BatchNormalization)   (None, None, None, 1 4096        conv_80[0][0]                 bnorm_92 (BatchNormalization)   (None, None, None, 5 2048        conv_92[0][0]                 bnorm_104 (BatchNormalization)  (None, None, None, 2 1024        conv_104[0][0]               leaky_80 (LeakyReLU)            (None, None, None, 1 0           bnorm_80[0][0]               leaky_92 (LeakyReLU)            (None, None, None, 5 0           bnorm_92[0][0]               leaky_104 (LeakyReLU)           (None, None, None, 2 0           bnorm_104[0][0]               conv_81 (Conv2D)                (None, None, None, 2 261375      leaky_80[0][0]               conv_93 (Conv2D)                (None, None, None, 2 130815      leaky_92[0][0]               conv_105 (Conv2D)               (None, None, None, 2 65535       leaky_104[0][0]               















YOLOv3 Architecture
YOLOv3 Architecture




Source: Ayoosh Kathuria, YOLOv3    
YOLO: https://pjreddie.com/yolo/













In [6]:



    


# YOLO parameters for image pre-processing and object detection

# set some parameters
net_h, net_w = 416, 416
obj_thresh, nms_thresh = 0.5, 0.45
anchors = [[116,90,  156,198,  373,326],  [30,61, 62,45,  59,119], [10,13,  16,30,  33,23]]

# YOLOv3 object classes: 80 MSCOCO classes
labels = ["person", "bicycle", "car", "motorbike", "aeroplane", "bus", "train", "truck",
  "boat", "traffic light", "fire hydrant", "stop sign", "parking meter", "bench",
  "bird", "cat", "dog", "horse", "sheep", "cow", "elephant", "bear", "zebra", "giraffe",
  "backpack", "umbrella", "handbag", "tie", "suitcase", "frisbee", "skis", "snowboard",
  "sports ball", "kite", "baseball bat", "baseball glove", "skateboard", "surfboard",
  "tennis racket", "bottle", "wine glass", "cup", "fork", "knife", "spoon", "bowl", "banana",
  "apple", "sandwich", "orange", "broccoli", "carrot", "hot dog", "pizza", "donut", "cake",
  "chair", "sofa", "pottedplant", "bed", "diningtable","toilet","tvmonitor","laptop","mouse",
  "remote", "keyboard", "cell phone", "microwave", "oven", "toaster", "sink", "refrigerator",
  "book", "clock", "vase", "scissors", "teddy bear", "hair drier", "toothbrush"]



    













In [7]:



    


# Sample images
image_path = './keras_yolo3_datasets/BCCD_Dataset/example.jpg'
image_path = './images/zebra.jpg'

image_dir1 = './data/images/'
image_fns1 = ['african_elephant.jpg', 'african_elephant_people.jpg', 'african_elephant_jeep', 
              'zebra.jpg', 'dog.jpg', 'eagle.jpg', 'giraffe.jpg', 'horses.jpg', 
              'kites_people.jpg', 'person.jpg']

# MS-COCO
image_dir2 = '../../datasets/mscoco/val2017/'
image_fns2 = ['000000000285.jpg', '000000000785.jpg', '000000004765.jpg', '000000014380.jpg']

# West Palm Beach
image_dir3 = './data/images/wpb/'
image_fns3 = ['wpb_marina_01.jpg', 'wpb_clematis_01.jpg']

# Art
image_dir4 = './data/images/art/'
image_fns4 = ['dama_sentada.jpg', 'scream.jpg']

image_path = image_dir1+image_fns1[1]
print("image_path:", image_path)



    















    






image_path: ./data/images/african_elephant_people.jpg





















YOLOv3 - image pre-processing
    

  • yolov3.preprocess_input()
    • 

  • Convert input image to defined width, height (416 x 416)
    • 














In [8]:



    


# YOLOv3 read & preprocess the sample image
image = cv2.imread(image_path)

image_h, image_w, _ = image.shape
new_image = xp_yolov3.preprocess_input(image, net_h, net_w)



    













In [9]:



    


# image, preprocessed new_image

print("image.shape:", image.shape, end=' - ')
print("new_image.shape:", new_image.shape)
plt.figure(figsize=(18,10))
plt.subplot(121)
plt.imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
plt.subplot(122)
plt.imshow(new_image[0])
plt.show();



    















    






image.shape: (1080, 1920, 3) - new_image.shape: (1, 416, 416, 3)










    


























In [10]:



    


# YOLOv3 helper functions

# Read and preprocess image
def yolov3_read_and_show(image_path):
    # image_path = image_dir4+image_fns4[0]
    image = cv2.imread(image_path)
    print("image_path:", image_path)
    print("image.shape:", image.shape)
    # Show image
    plt.figure(figsize=(18,10))
    plt.imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    plt.show();
    return image

# Object detection and display of annotated image
def yolov3_detect_and_show(image, image_path):
    print("YOLOv3 predicting image_path:", image_path)
    image_h, image_w, _ = image.shape
    new_image = xp_yolov3.preprocess_input(image, net_h, net_w)
    yolos = yolov3.predict(new_image)
    boxes = []
    for i in range(len(yolos)):
        boxes += xp_yolov3.decode_netout(yolos[i][0], anchors[i], 
                                         obj_thresh, nms_thresh, net_h, net_w)
    xp_yolov3.correct_yolo_boxes(boxes, image_h, image_w, net_h, net_w)
    xp_yolov3.do_nms(boxes, nms_thresh)     
    xp_yolov3.draw_boxes(image, boxes, labels, obj_thresh) 
    detected_image_path = image_path[:-4] + '_detected' + image_path[-4:]
    cv2.imwrite(detected_image_path, (image).astype('uint8')) 
    # Show detected bounding boxes on image
    image = cv2.imread(detected_image_path)
    plt.figure(figsize=(18,10))
    plt.imshow(cv2.cvtColor(image, cv2.COLOR_BGR2RGB))
    plt.show()
    return boxes

# Print object predictions and detected image; Multi-class, multi-object proposal per box
def print_predictions(boxes, labels, obj_thresh):
    print("Predictions on image (object threshold: {})".format(obj_thresh))
    obj_num = 0
    for b, box in enumerate(boxes):
        label_str = ''
        label = -1
        for i in range(len(labels)):
            if box.classes[i] > obj_thresh:
                obj_num += 1
                label_str += labels[i]
                label = i
                info_str = 'object ' + str(obj_num) + ' box ' + str(b)
                info_str += ' - ' + labels[i] + ' (class ' + str(i) + '): '
                print(info_str + str(box.classes[i]*100) + '%')
    return obj_num



    















YOLOv3 Object Detection - Elephant and people












In [11]:



    


# Elephant image
image_path = image_dir1+image_fns1[1]
# Object detection
image = yolov3_read_and_show(image_path)



    















    






image_path: ./data/images/african_elephant_people.jpg
image.shape: (1080, 1920, 3)










    


























In [12]:



    


# Object detection
boxes = yolov3_detect_and_show(image, image_path)



    















    






YOLOv3 predicting image_path: ./data/images/african_elephant_people.jpg
elephant: 99.82871413230896%
person: 99.97822642326355%
person: 99.86847043037415%










    


























In [13]:



    


# Print yolo boxes and classes
objects_total = print_predictions(boxes, labels, obj_thresh)
print("\nobjects_total:", objects_total)



    















    






Predictions on image (object threshold: 0.5)
object 1 box 253 - elephant (class 20): 99.82871413230896%
object 2 box 1724 - person (class 0): 99.97822642326355%
object 3 box 1736 - person (class 0): 99.86847043037415%

objects_total: 3





















YOLOv3 Object Detection - West Palm Beach












In [14]:



    


# WPB downtown image
image_path = image_dir3+image_fns3[1]
image = yolov3_read_and_show(image_path)



    















    






image_path: ./data/images/wpb/wpb_clematis_01.jpg
image.shape: (500, 800, 3)










    


























In [15]:



    


# Object detection
boxes = yolov3_detect_and_show(image, image_path)



    















    






YOLOv3 predicting image_path: ./data/images/wpb/wpb_clematis_01.jpg
bus: 99.76127743721008%
person: 99.40769672393799%
person: 95.56765556335449%
bicycle: 99.69530701637268%
traffic light: 87.71517276763916%
traffic light: 84.3330442905426%
bus: 63.52843642234802%
car: 68.71739029884338%
car: 97.94415235519409%
car: 91.70721173286438%
motorbike: 67.31112599372864%










    


























In [16]:



    


# Print yolo boxes and classes
objects_total = print_predictions(boxes, labels, obj_thresh)
print("\nobjects_total:", objects_total)



    















    






Predictions on image (object threshold: 0.5)
object 1 box 282 - bus (class 5): 99.76127743721008%
object 2 box 1878 - person (class 0): 99.40769672393799%
object 3 box 1893 - person (class 0): 95.56765556335449%
object 4 box 1957 - bicycle (class 1): 99.69530701637268%
object 5 box 4661 - traffic light (class 9): 87.71517276763916%
object 6 box 4847 - traffic light (class 9): 84.3330442905426%
object 7 box 6755 - bus (class 5): 63.52843642234802%
object 8 box 7078 - car (class 2): 68.71739029884338%
object 9 box 7081 - car (class 2): 97.94415235519409%
object 10 box 7286 - car (class 2): 91.70721173286438%
object 11 box 7454 - motorbike (class 3): 67.31112599372864%

objects_total: 11





















YOLOv3 Object Detection - Dama sentada 1937












In [17]:



    


# Dama image (Picasso)
image_path = image_dir4+image_fns4[0]
# Object detection
image = yolov3_read_and_show(image_path)



    















    






image_path: ./data/images/art/dama_sentada.jpg
image.shape: (3520, 2666, 3)










    


























In [18]:



    


# Object detection
boxes = yolov3_detect_and_show(image, image_path)



    















    






YOLOv3 predicting image_path: ./data/images/art/dama_sentada.jpg
person: 80.92698454856873%
chair: 53.65053415298462%










    


























In [19]:



    


# Print yolo boxes and classes
objects_total = print_predictions(boxes, labels, obj_thresh)
print("\nobjects_total:", objects_total)



    















    






Predictions on image (object threshold: 0.5)
object 1 box 296 - person (class 0): 80.92698454856873%
object 2 box 371 - chair (class 56): 53.65053415298462%

objects_total: 2





















Run YOLOv3 Object Detection - Dog












In [20]:



    


# Dog image (Redmon)
image_path = image_dir1+image_fns1[4]
# Object detection
image = yolov3_read_and_show(image_path)



    















    






image_path: ./data/images/dog.jpg
image.shape: (576, 768, 3)










    


























In [21]:



    


# Object detection
boxes = yolov3_detect_and_show(image, image_path)



    















    






YOLOv3 predicting image_path: ./data/images/dog.jpg
bicycle: 99.34678077697754%
dog: 98.6376702785492%
truck: 92.89805889129639%










    




























Notes
YOLOv3 Training is done in a separate Jupyter notebook.


    

  • See: yolov3_object_detector_train.ipynb.
    • 





YOLOv3 model trained using the DarkNet code base on the MSCOCO dataset.


Output of the model:


    

  • encoded candidate bounding boxes from three different grid scale sizes
    • 

  • bounding boxes defined in the context of anchor boxes, 
    
chosen based on an analysis of the size of objects in the MSCOCO dataset.
    • 














In [ ]: